Factor and figure management

Initialize the data

  • Load the required libraries:
suppressPackageStartupMessages(library(tidyverse))
suppressPackageStartupMessages(library(gapminder))
suppressPackageStartupMessages(library(knitr))
suppressPackageStartupMessages(library(plotly))
suppressPackageStartupMessages(library(broom))
suppressPackageStartupMessages(library(scales))
  • We’ll use forcats to help re-order factors (package located inside the tidyverse) and plotly to enhance the plot visualization. Broom is handy for providing statistical analyses.
  • We will check the structure of the data first to sanity check that we are working with factors:
gapminder %>% str()
## Classes 'tbl_df', 'tbl' and 'data.frame':    1704 obs. of  6 variables:
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ pop      : int  8425333 9240934 10267083 11537966 13079460 14880372 12881816 13867957 16317921 22227415 ...
##  $ gdpPercap: num  779 821 853 836 740 ...
  • We’ve confirmed that country is a factor with 142 levels and continent is a factor with 5 levels. We’ll proceed with manipulating the data.

Part 1 - Factor Management

Characterize data before and after re-levelling

Drop Oceania. Filter the Gapminder data to remove observations associated with the continent of Oceania. Additionally, remove unused factor levels. Provide concrete information on the data before and after removing these rows and Oceania; address the number of rows and the levels of the affected factors.

Reorder the levels of country or continent. Use the forcats package to change the order of the factor levels, based on a principled summary of one of the quantitative variables. Consider experimenting with a summary statistic beyond the most basic choice of the median.

Method

  • Drop Oceania by filtering the data to remove observations
  • Remove unused factor levels
  • Provide concrete information on the data before/after the manipulation (e.g. effect on number of rows/levels of affected factors)
  • Re-order the levels of continent

Code

First, we will examine a summary of the initial data:

gapminder %>% 
  summary() %>% #Evaluate the factors and observations per factor
  kable() #Enhance the table output
country continent year lifeExp pop gdpPercap
Afghanistan: 12 Africa :624 Min. :1952 Min. :23.60 Min. :6.001e+04 Min. : 241.2
Albania : 12 Americas:300 1st Qu.:1966 1st Qu.:48.20 1st Qu.:2.794e+06 1st Qu.: 1202.1
Algeria : 12 Asia :396 Median :1980 Median :60.71 Median :7.024e+06 Median : 3531.8
Angola : 12 Europe :360 Mean :1980 Mean :59.47 Mean :2.960e+07 Mean : 7215.3
Argentina : 12 Oceania : 24 3rd Qu.:1993 3rd Qu.:70.85 3rd Qu.:1.959e+07 3rd Qu.: 9325.5
Australia : 12 NA Max. :2007 Max. :82.60 Max. :1.319e+09 Max. :113523.1
(Other) :1632 NA NA NA NA NA

Oceania has 24 observations in the original data set. Next we will drop the observations related to Oceania from the data set:

gapminder_dropOc <- gapminder %>%
  filter(continent != "Oceania") 

gapminder_dropOc %>% 
  summary() %>%  
  kable()
country continent year lifeExp pop gdpPercap
Afghanistan: 12 Africa :624 Min. :1952 Min. :23.60 Min. :6.001e+04 Min. : 241.2
Albania : 12 Americas:300 1st Qu.:1966 1st Qu.:48.08 1st Qu.:2.780e+06 1st Qu.: 1189.1
Algeria : 12 Asia :396 Median :1980 Median :60.34 Median :7.024e+06 Median : 3449.5
Angola : 12 Europe :360 Mean :1980 Mean :59.26 Mean :2.990e+07 Mean : 7052.4
Argentina : 12 Oceania : 0 3rd Qu.:1993 3rd Qu.:70.75 3rd Qu.:1.987e+07 3rd Qu.: 8943.2
Austria : 12 NA Max. :2007 Max. :82.60 Max. :1.319e+09 Max. :113523.1
(Other) :1608 NA NA NA NA NA 
gapminder_dropOc %>% 
  str()
## Classes 'tbl_df', 'tbl' and 'data.frame':    1680 obs. of  6 variables:
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ pop      : int  8425333 9240934 10267083 11537966 13079460 14880372 12881816 13867957 16317921 22227415 ...
##  $ gdpPercap: num  779 821 853 836 740 ...

Oceania now has 0 observations, however it is still present in the data frame as a factor. Next, we will drop Oceania as an unused factor using the droplevels() function from the forcats package:

gapminder_dropOc <- gapminder_dropOc %>%
  droplevels()

gapminder_dropOc %>% 
  summary() %>% 
  kable()
country continent year lifeExp pop gdpPercap
Afghanistan: 12 Africa :624 Min. :1952 Min. :23.60 Min. :6.001e+04 Min. : 241.2
Albania : 12 Americas:300 1st Qu.:1966 1st Qu.:48.08 1st Qu.:2.780e+06 1st Qu.: 1189.1
Algeria : 12 Asia :396 Median :1980 Median :60.34 Median :7.024e+06 Median : 3449.5
Angola : 12 Europe :360 Mean :1980 Mean :59.26 Mean :2.990e+07 Mean : 7052.4
Argentina : 12 NA 3rd Qu.:1993 3rd Qu.:70.75 3rd Qu.:1.987e+07 3rd Qu.: 8943.2
Austria : 12 NA Max. :2007 Max. :82.60 Max. :1.319e+09 Max. :113523.1
(Other) :1608 NA NA NA NA NA 
gapminder_dropOc %>% 
  str()
## Classes 'tbl_df', 'tbl' and 'data.frame':    1680 obs. of  6 variables:
##  $ country  : Factor w/ 140 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ continent: Factor w/ 4 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ pop      : int  8425333 9240934 10267083 11537966 13079460 14880372 12881816 13867957 16317921 22227415 ...
##  $ gdpPercap: num  779 821 853 836 740 ...

We observe that Oceania has now disappeared from the continent list. Further, we see that the factor continent now has only 4 levels. The original data set had 1704 observations of 6 variables (i.e. 1704 rows and 6 columns), while the manipulated data set has 1680 observations of 6 variables. Therefore we observe a concrete reduction on the data. When manipulating a data set through filters, it is advantageous to create a new variable for the manipulated data in order to refer to it in the future and maintain the integrity of the original data set.

  • Next we will create a principled summary of the data based on the quantitative variable gdpPercap. We want to evaluate the rank of countries in Europe based on gdpPercap. We will start by plotting the data as is:
Europe_gdp <- gapminder %>% 
  select(continent, country, gdpPercap,year) %>%  #Reduce the size of the data set for faster processing
  filter(continent == "Europe") 

Europe_gdp %>% 
  str()
## Classes 'tbl_df', 'tbl' and 'data.frame':    360 obs. of  4 variables:
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 4 4 4 4 4 4 4 4 4 4 ...
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 2 2 2 2 2 2 2 2 2 2 ...
##  $ gdpPercap: num  1601 1942 2313 2760 3313 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
Europe_gdp %>% 
  head() %>% 
  kable()
continent country gdpPercap year
Europe Albania 1601.056 1952
Europe Albania 1942.284 1957
Europe Albania 2312.889 1962
Europe Albania 2760.197 1967
Europe Albania 3313.422 1972
Europe Albania 3533.004 1977 
Europe_gdp %>% 
  ggplot(aes(country, gdpPercap)) + 
  geom_violin() +
  labs(title = "GDP per capita - Europe",
    x = "Country", y = "GDP per capita") +
  scale_y_log10(labels=dollar_format()) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))  #Rotate x labels

Plotting the data directly doesn’t provide any insight into a correlation between these two variables. We will next arrange by gdpPerCap to see whether this has an effect on the table and plot:

Europe_gdp_arr <- Europe_gdp %>% 
  arrange(gdpPercap)

Europe_gdp_arr %>% 
  str()
## Classes 'tbl_df', 'tbl' and 'data.frame':    360 obs. of  4 variables:
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 4 4 4 4 4 4 4 4 4 4 ...
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 13 13 2 13 2 132 13 132 2 132 ...
##  $ gdpPercap: num  974 1354 1601 1710 1942 ...
##  $ year     : int  1952 1957 1952 1962 1957 1952 1967 1957 1962 1962 ...
Europe_gdp_arr %>% 
  head() %>% 
  kable()
continent country gdpPercap year
Europe Bosnia and Herzegovina 973.5332 1952
Europe Bosnia and Herzegovina 1353.9892 1957
Europe Albania 1601.0561 1952
Europe Bosnia and Herzegovina 1709.6837 1962
Europe Albania 1942.2842 1957
Europe Turkey 1969.1010 1952 
Europe_gdp_arr %>% 
  ggplot(aes(country, gdpPercap)) + 
  geom_violin() +
  labs(title = "GDP per capita - Europe",
    x = "Country", y = "GDP per capita") +
  scale_y_log10(labels=dollar_format()) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))  #Rotate x labels

We observe the arrange did not affect the structure or the plot, however the table output is clearly different. Next we will evaluate the effect of using the forcats package to re-order the data:

Europe_gdp %>% 
  ggplot(aes(fct_reorder(country, gdpPercap), gdpPercap)) + 
  geom_violin() +
  labs(title = "GDP per capita - Europe",
    x = "Country", y = "GDP per capita") +
  scale_y_log10(labels=dollar_format()) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))  #Rotate x labels

We observe that fct_reorder does affect the plot, as it now shows countries in order of ascending median GDP per capita. Finally, we will examine the effect of combining arrange and fct_reorder:

Europe_gdp_arr %>% 
  ggplot(aes(fct_reorder(country, gdpPercap), gdpPercap)) + 
  geom_violin() +
  labs(title = "GDP per capita - Europe",
    x = "Country", y = "GDP per capita") +
  scale_y_log10(labels=dollar_format()) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))  #Rotate x labels

Thus, coupling arrange and fct_reorder allows us to manipulate the structure, table and plot, so it is more comprehensive to use both commands if all three types of output are desired.

Part 2 - File I/O

Experiment with one or more of write_csv()/read_csv() (and/or TSV friends), saveRDS()/readRDS(), dput()/dget(). Create something new, probably by filtering or grouped-summarization of Singer or Gapminder. I highly recommend you fiddle with the factor levels, i.e. make them non-alphabetical (see previous section). Explore whether this survives the round trip of writing to file then reading back in.

Method

  • Create a new data set
  • Arrange the data set
  • Write to file
  • Read from file

Code

gapminder_lifeExp <- gapminder %>% 
  select(lifeExp, continent, year) %>% 
  group_by(continent, year) %>% 
  summarise(mean_gdp = mean(lifeExp))

gapminder_lifeExp %>% 
  head() %>% 
  kable()
continent year mean_gdp
Africa 1952 39.13550
Africa 1957 41.26635
Africa 1962 43.31944
Africa 1967 45.33454
Africa 1972 47.45094
Africa 1977 49.58042 
gapminder_lifeExp_arr <- gapminder_lifeExp %>% 
  arrange(mean_gdp)

gapminder_lifeExp_arr %>% 
  head() %>% 
  kable()
continent year mean_gdp
Africa 1952 39.13550
Africa 1957 41.26635
Africa 1962 43.31944
Africa 1967 45.33454
Asia 1952 46.31439
Africa 1972 47.45094

We observe the difference between the tables after the arrange function has been applied. Now we will test if the arrangment is preserved after writing and reading to a file:

write_csv(gapminder_lifeExp_arr, "gapminder_lifeExp_arr.csv")

read_csv("gapminder_lifeExp_arr.csv") %>% 
  head() %>% 
  kable()
## Parsed with column specification:
## cols(
##   continent = col_character(),
##   year = col_integer(),
##   mean_gdp = col_double()
## )
continent year mean_gdp
Africa 1952 39.13550
Africa 1957 41.26635
Africa 1962 43.31944
Africa 1967 45.33454
Asia 1952 46.31439
Africa 1972 47.45094

The new arrangement of the data was preserved, therefore we learned that arrange can be used to manipulate tables that are written externally to files, in addition to tables internally within the R environment, as we learned in the previous section.

Part 3 - Visualization Design

Next I will improve my first plot using techniques from visualization design to enhance the appearance and readability. My first plot looked like this:

Europe_gdp_arr %>% 
  ggplot(aes(fct_reorder(country, gdpPercap), gdpPercap)) + 
  geom_violin() +
  labs(title = "GDP per capita (European countries)",
    x = "Country", y = "GDP per capita") +
  scale_y_log10() +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))  #Rotate x labels

I applied a few different techniques from class to improve the plot visually: * Changing the scale to dollar_format() * Using the black & white theme * Eliminating the unnecessary borders on the top and right side of the plot * Centering the title * Applying a continuous colour scheme to visually separate the countries by colour * Removing the legend that is generated from applying the colour scheme

Overall, I believe that the plot is visually easier to read and more appealing to the viewer, due to the removal of unnecessary features and the addition of colour:

Europe_gdp_plot <- Europe_gdp_arr %>% 
  ggplot(aes(fct_reorder(country, gdpPercap), gdpPercap, fill = country), alpha = 0.2) + 
  geom_violin() +
  labs(title = "GDP per capita (European countries)",
    x = "Country", y = "GDP per capita") +
  scale_y_log10(labels=dollar_format()) +
  theme_bw() +
  theme(panel.border = element_blank(), 
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(), 
       axis.line = element_line(colour = "black")) +
  theme(plot.title = element_text(hjust = 0.5)) +
  guides(fill = FALSE) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1, size = 10))

Europe_gdp_plot

Then, we will convert this plot into a plotly plot using the porting function in ggplotly:

Europe_gdp_plot %>% 
  ggplotly()

The biggest advantage that I can glean from using plotly is the ability to read data point values using the cursor. By mousing over any location on the graph, it is possible to see the exact (x,y) coordinates. The conversion also added the legend back in, but the rest of the formatting was preserved. The overal aesthetic is slightly cleaner and crisper than the **ggplot* plot. I believe this is due to the change in font and crisper borders.

Plotly can also be used to create 3D plots and HTML widgets, which can’t be done in ggplot:

plot_ly(
  Europe_gdp_arr, 
  x = ~country, 
  y = ~gdpPercap, 
  z = ~year,
  type = "scatter3d",
  mode = "markers",
  opacity = 0.5
  ) #%>% 
  #htmlwidgets::saveWidget("plotly.html"

Part 4 - Writing figures to file

Use ggsave() to explicitly save a plot to file. Then use dfdf to load and embed it in your report. You can play around with various options, such as:

Arguments of ggsave(), such as width, height, resolution or text scaling. Various graphics devices, e.g. a vector vs. raster format. Explicit provision of the plot object p via ggsave(…, plot = p). Show a situation in which this actually matters.

Finally, we will save the plot as a png image.

ggsave("Europe_gdp_plot.png", plot = Europe_gdp_plot, width = 18, height = 8, units = "cm")

The plot image is located here